A two-step approach to optimizing meshes

Meshes obtained from three-dimensional scanning usually contain huge amounts of noise. It is therefore necessary to optimize them before further applications. This paper proposes a novel approach to optimize arbitrary meshes with feature preservation, which consists of two steps. First, a modified bilateral filtering is employed to denoise the meshes, which can be considered as an iterative, local, non-linear, and anisotropic mesh optimization. Then the resulting meshes is smoothed by solving a Laplacian equation, which is regarded as a non-iterative, global, linear, and isotropic mesh optimization. The performance of our approach is experimentally validated via comparisons with conventional techniques. The results demonstrate our approach successfully removes noise while faithfully preserving the original surface geometry, and avoids volume shrinkage and shape distortion.